Antarctic Surface Hydrology and Impacts on Ice-Sheet Mass Balance

Home , Law Glacier, Moulin (geomorphology)

Perspective https://doi.org/10.1038/s41558-018-0326-3

Antarctic surface hydrology and impacts on ice-sheet mass balance

Robin E. Bell 1*, Alison F. Banwell 2,3, Luke D. Trusel 4 and Jonathan Kingslake 1,5

Melting is pervasive along the ice surrounding Antarctica. On the surface of the grounded ice sheet and floating ice shelves, extensive networks of lakes, streams and rivers both store and transport water. As melting increases with a warming climate, the surface hydrology of Antarctica in some regions could resemble Greenland’s present-day ablation and percolation zones. Drawing on observations of widespread surface water in Antarctica and decades of study in Greenland, we consider three modes by which meltwater could impact Antarctic mass balance: increased runoff, meltwater injection to the bed and meltwater-induced ice-shelf fracture — all of which may contribute to future ice-sheet mass loss from Antarctica.

urface meltwater in Antarctica is more extensive than previ- Through-ice fractures are interpreted as evidence of water having ously thought and its role in projections of future mass loss drained through ice shelves (Figs 1a and 2)15. Similar to terrestrial Sare becoming increasingly important. As accurately projecting hydrologic systems, these components of the Antarctic hydrologic future sea-level rise is essential for coastal communities around the system store, transport and export water. In contrast to terrestrial globe, understanding how surface melt may either trigger or buf- hydrology, on ice sheets and glaciers water can refreeze with conse- fer rapid changes in ice flow into the ocean is critical. We provide quences for the temperature of the surrounding ice16–18, firn or snow. an overview of the current understanding of the major components Storage occurs in lakes, crevasses, in buried lakes and possibly in of the Antarctic surface hydrology system and the distribution of firn aquifers. Transport and export are less persistent and more dif- melt. Using the framework of surface hydrology in Greenland, we ficult to observe than lakes14,19,20. Antarctic surface and subsurface consider the different ways in which surface hydrology can impact hydrological systems have been studied using satellite and airborne ice-sheet mass balance. Looking to the future, we discuss how the imagery1,8,11–14, although field-based observations are limited8,21–23. hydrologic systems will evolve in Antarctica as well as their impact on future changes in ice-sheet mass balance. Finally, we highlight Surface storage of meltwater. Meltwater is stored in surface lakes on knowledge gaps that limit our understanding of the impact of both grounded and floating ice. On grounded ice, lakes develop in increased surface meltwater on future sea level rise. areas with local-scale melt enhancement and relatively low accumulation rates; areas that are often close to rock outcrops and blue ice Current distribution of meltwater (for example, Shackleton Glacier; Fig. 1j)12,14. Similar to Greenland20, Meltwater on the surface of Antarctica was observed by early explor- on grounded Antarctic ice, lakes form in persistent surface depres- ers, who noted the noise of running water and water seeping into sions. The formation of surface lakes in the same location each their tents1. Today the surface melt distribution in Antarctica (Fig. 1) year over decades is evidence of control by the interplay between is determined using satellite observations2–5 and reanalysis-forced bedrock topography and ice flow24. On the floating ice shelves sur- regional climate modeling6. The surface meltwater production esti- rounding Antarctica8,14,19,20,25, water collects in surface depressions mates derived from these two methods correspond well with in situ that move with ice flow. These surface depressions in the ice shelf observations7. At present, the most intense melt is observed across are controlled by basal crevassing26, grounding zone flow-stripe ice shelves (Fig. 1), particularly along the Antarctic Peninsula, development27, suture-zone depressions1 and basal channels pro- including the Larsen C, Wilkins and George VI ice shelves, as well duced by ocean melting28. Water will fill a depression if the ice sur- as the relatively low-latitude East Antarctic ice shelves, including the face or near surface is impermeable. Impermeable surfaces are often West and Shackleton ice shelves. More localized, but intense, melt associated with high melt and low snow accumulation rates29. Once occurs on other East Antarctic ice shelves, including the Amery and water collects in an ice shelf depression, the basin will deepen due to Roi Baudouin ice shelves7,8, where extensive surface hydrological both enhanced lake-bottom ablation (due to the lower albedo of the networks develop. The two largest ice shelves, Ross and Ronne- water compared to the surrounding ice/snow)30,31, and the flexural Filchner, experience only minor surface melting. The upper eleva- response of the floating ice to the water load32,33. The largest supra- tion limit of surface melting today is generally ~1,400 m during glacial lake (~80 km long) is on the Amery Ice Shelf (Fig. 2e)13,14,34. spatially extensive, but low-magnitude, West Antarctic melt epi- On both grounded and floating ice, surface fractures (crevasses) sodes7,9, compared to a 3,200 m elevation limit in Greenland during can accumulate water26, serving as another storage site for meltwa- the anomalous10 2012 melt events. ter and a mechanism by which water directly impacts ice dynam- Liquid water on the Antarctic Ice Sheet and the floating ice ics. Water-filled fractures may propagate vertically when sufficient shelves that buttress upstream grounded ice (Fig. 1) is found in water is available, creating through-ice fractures on Antarctica’s ice supraglacial lakes, subsurface lakes, surface streams and rivers1,8,11–14. shelves32,35,36 and Greenland’s floating tongues37. Fractures beneath

1Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA. 2Scott Polar Research Institute, University of Cambridge, Cambridge, UK. 3Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, CO, USA. 4Department of Geology, Rowan University, Glassboro, NJ, USA. 5Department of Earth and Environmental Sciences, Columbia University, Palisades, NY, USA. *e-mail: [email protected]

1044 Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange NAture CliMAte ChAnGe Perspective

a b c Meltponds and dolines Foehn-enhanced melt ponds Subsurface lake

20 km 20 km George VI Ice Shelf Larsen C Ice Shelf Western Roi Baudouin Ice Shelf 10 January 2003 7 February 2016 January/February 2016 j d Southernmost large-scale drainage Ice shelf stream and moulin c d

Antarctic Peninsula Dronning Maud Land b

a e Shackleton Glacier Eastern Roi Baudouin Ice Shelf 10 January 2010 January/February 2016 e i West Antarctica East Antarctica High-elevation stream Largest supraglacial lake i f j h g Surface melt (mm w.e. yr–1) 80 km 0 100 200 300 400 500 Law Glacier (~1,830 m a.s.l.) Amery Ice Shelf 14 January 2017 21 February 1988 h g f Drainage across grounding line Terminal waterfall Ring and radial fractures, streams and ponds

10 km

Darwin Glacier/Ross Ice Shelf Nansen Ice Shelf Shackleton Ice Shelf 31 December 2001 12 January 2014 10 February 1947

Fig. 1 | Examples of major components of surface hydrological systems located on a present-day Antarctic surface melt map. The central map shows 2000–2009 Antarctica surface melt from QuikSCAT satellite observation 7; the locations of the images in a–j are indicated. a, Meltwater lakes and dolines (arrows), b, Foehn wind-enhanced meltwater ponding. c, Buried lake. d, Moulin draining surface stream. e, Elongate supraglacial lake. f, Fractures around a drained lake. Scale unknown. g, Persistent waterfall draining water. h, Supraglacial streams transporting water across grounding line of the Darwin Glacier onto the Ross Ice Shelf. i, High-elevation (1,830 m) meltwater stream. j, Meltwater stream crossing the grounding line. Images reproduced from: US Geological Survey (a,b,e,h); ref. 8, Springer Nature Limited (c); Sanne Bosteels (d); USGS/EROS and the Polar Geospatial Center (f); Won Sang Lee (g); Mike Kaplan (i); John Stone (j).

lakes on Greenland’s grounded ice drain meltwater to the ice-sheet line lakes refreeze, they form massive ice layers29,42. Over succes- bed by hydrofracturing38,39. There is no direct evidence of hydro- sive melt seasons, frozen surface lakes (now stacked ice lenses) may fracture beneath lakes on grounded Antarctic ice. accumulate in dense and thick ice horizons29. On the Larsen C Ice Shelf, a massive ice facies >40 m thick, extending 16 km horizon- Englacial storage of meltwater. Antarctic surface meltwater tally, was interpreted as a stack of frozen lakes42. Temperature pro- is stored englacially when surface lakes freeze over and are bur- files through this refrozen ice are substantially warmer due to the ied by snowfall8,40. In Antarctica, buried lakes tend to form on ice release of latent heat as the lakes froze, similar to the cryohydro- shelves close to the grounding line8. Since at least 1947 on the Roi logic warming described across Greenland17. Baudouin Ice Shelf, meltwater produced in areas of blue ice above In Greenland, perennial firn aquifers store water in environ- and below the grounding line fills surface lakes. These lakes are ments similar to where buried supraglacial lakes form43. Water in buried as the ice moves towards the calving front14. Radar satel- these firn aquifers is stored in a porous matrix of ice crystals. No lites, such as C-Band Sentinel-1 A and B, are capable of penetrating Antarctic firn aquifer has been sampled yet, but beneath massive metres though dry snow, highlighting the promise of tracing buried ice facies on Larsen C, a second ~45-m-thick ice unit has been lakes and other subsurface liquid water41. When these grounding interpreted as a percolation-type facies of water-infiltrated firn42.

Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange 1045 Perspective NAture CliMAte ChAnGe

19 Nunatak the Larsen B Ice Shelf before its collapse (T. A. Scambos, personal Blue ice communication). Simple routing calculations indicate that meltwater could be removed from other Antarctic ice shelves such as Ross, Amery, Filchner Ronne and Larsen C1. Transport of meltwater off Streams floating ice shelves has the potential to buffer ice shelves from frac- Grounding line ture and collapse associated with surface lakes. Lakes Drivers of meltwater distribution River Doline Fractures Antarctic surface meltwater distribution is driven at present by Waterfall regional shifts in climate together with the influence of local- scale process and microclimates. The predominance of melting on Antarctic Peninsula ice shelves today reflects the rapid regional Icebergs 51 Possible lake with atmospheric warming that began in the 1950s . The resulting melt basal fracture intensification on ice shelves is thought to be directly responsible for draining to bed 52,53 Aquifer/buried multiple ice-shelf collapses over recent decades . These collapses, lake together with the associated loss of buttressing, have triggered Antarctic Peninsula outlet-glacier acceleration54. An ice core from James Ross Island on the northeast Antarctic Peninsula indicates that surface melting rapidly increased in the late twentieth century Fig. 2 | Antarctic surface hydrology. The major components of the modern relative to the past 1,000 years55. Observed warming and melt inten- Antarctic hydrologic system are shown. The possible future surface-to- sification across the northeastern Antarctic Peninsula are associated bed connection is included, illustrated as a lake-bottom fracture draining with a strengthening of the circumpolar westerly winds marked by meltwater to the base of the ice sheet, based on Greenland analogues. the positive phase shift in the Southern Annular Mode since the Dolines are locally uplifted, empty depressions, interpreted as evidence 56 15 1970s , which in turn is considered to be the result of coincident of surface lakes that have drained through ice shelves via ice fractures . anthropogenically induced depletion of stratospheric ozone57. Nunataks are areas of exposed rock on the grounded ice. Broader-scale climate dynamics also impact Antarctic surface melting, including oceanic–atmospheric variability in the tropical Pacific58,59. Striking examples of this linkage are anomalous, exten- In Antarctica, drainage systems often terminate where they deliver sive melt events across the Ross Ice Shelf and the West Antarctic water into snow-covered areas1,8,12,14. Perennial firn aquifers could Ice Sheet that have been linked to an El Niño/Southern Oscillation develop at these sites if accumulation rates are sufficiently high to (ENSO) teleconnection pattern that favours warm, marine air insulate the downward-percolating liquid water from low winter- intrusions into West Antarctica5,9,60. Antarctic climate and surface time surface temperatures, or if the water is routed deep enough to melting are strongly coupled to broader climate system dynamics be thermally isolated from the surface. Perennial firn aquifers occur and anthropogenic forcing. in Greenland17,44–48 in locations with both moderate and high melt Local-scale processes also drive the distribution of Antarctic rates (>650 mm of water equivalent (w.e.) yr−1)44 and high snow surface melt. Exposure of low-albedo blue ice and bedrock near accumulation rates (~1–5 m w.e. yr−1)45. Similar high snow accu- ice-shelf grounding zones can enhance melting through a positive mulation rates occur today on the western Antarctic Peninsula49, melt–albedo feedback8,14. On the ice sheet, blue ice areas gener- as well as on the upwind flanks of coastal domes and the ice-sheet ally produce greater meltwater volumes than the adjacent snow- margins of West Antarctica, but surface melt rates are currently low covered regions. As blue ice22 only covers 1.6% of the surface of in these regions. Antarctica61,62, the overall volume of meltwater produced by local- scale melt enhancement over blue ice areas is thought to be a small Surface meltwater transport. Across broad sectors of Antarctica, fraction of the Antarctic surface melt. Observations and modelling meltwater transport over the surface of the ice sheet and ice shelves of meltwater production across ice-covered areas are particularly occurs along relatively low surface slopes through networks of streams lacking in Antarctica. and rivers. In some cases, water moves tens to hundreds of kilome- Winds play an important role in surface meltwater production tres14 and has persisted for decades. The Transantarctic Mountains across Antarctica. Warming of descending katabatic winds that support some of the continent’s most high-latitude (~85° S) persistently drain from the Antarctic interior, and associated wind and high-elevation (~1,800 m a.s.l.) meltwater drainage systems scouring and blue ice exposure are known to locally enhance surface (Fig. 1). It is currently unclear how melting in these extreme loca- melting across ice-shelf grounding zones in Dronning Maud Land, tions supports these persistent drainage systems, but it is presum- East Antarctica8. Analogous processes enhance melt on the Ross Ice ably related to the abundance of low-albedo bedrock and downslope Shelf, as well on the innermost Amery Ice Shelf3,5. Foehn winds play winds that emanate from the East Antarctic plateau. Streams and a similar role in melt generation. Although more episodic and less rivers may affect ice-sheet mass balance by moving water onto ice directionally constant than katabatics, warm, dry and clear sky con- shelves where ponding water can contribute to ice-shelf collapse. ditions associated with foehn wind events enhance melting across Meltwater streams feed lakes in high-albedo snow on the Riiser- eastern Antarctic Peninsula ice shelves6,63,64 and the McMurdo Dry Larsen, Amery, Nivlisen and Roi Baudouin ice shelves8,14,50. Ice Valleys65,66. Local melt enhancement produced by foehn winds shelves receiving meltwater through this mechansim are more likey is linked to depletion of ice shelf firn pore space67 and meltwater to affect ice-sheet mass balance if they are both suspectible to frac- ponding on innermost Larsen C Ice Shelf7,42,63. As firn air depletion turing and buttress large upstream catchments. results in an impermeable ice surface, this process is an important Streams and rivers can also transport meltwater off ice shelves in precursor for meltwater-induced hydrofracture6,68. Foehn winds the ocean via waterfalls1 at the calving ice front, or through moulins, probably contributed to the collapse of the Larsen B Ice Shelf69. A dolines and crevasses12,19. On the Nansen Ice Shelf1, a waterfall fed by result of the interplay of Antarctic topography and prevailing winds, a surface river has persisted since at least 1974. This river and water- wind-enhanced melting will continue to be an important compo- fall system drains a significant fraction of the meltwater formed on nent of Antarctic surface meltwater production and hydrology in the ice shelf into the Ross Sea. Similar water export was observed on coming decades.

1046 Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange NAture CliMAte ChAnGe Perspective

Modes of meltwater impact As Antarctic climate warming results in the development on grounded Surface meltwater on ice sheets and the adjacent floating ice shelves ice of more extensive surface lakes, aquifers and rivers, in some areas has the potential to significantly impact ice-sheet mass balance. We the surface and basal systems may connect. We suggest that a switch focus on three primary modes of meltwater influence on ice-sheet from an ice-sheet base that is isolated from surface melt to one that mass balance: (1) surface melt leading to direct surface runoff and receives seasonal injections of surface meltwater could trigger a fun- thinning (Fig. 3a,b); (2) changing the basal thermal and hydrologi- damental shift in the dynamics and mass balance of Antarctica. cal state by injection of surface meltwater into the subglacial environment (Fig. 3c,d); and (3) meltwater-induced ice-shelf collapse Meltwater-induced ice-shelf collapse. The third mode (Mode 3 in (Fig. 3e,f), producing an acceleration of mass loss from the Fig. 3e,f), is active today in Antarctica. Through-ice fractures on ice upstream outlet glaciers. Other influences of surface meltwater shelves may develop via two mechanisms: the downward propaga- include cryohydrologic warming and enhanced ocean melting70,71. tion of water-filled fractures68, referred to as hydrofracture35, and Cryohydrologic warming in a lake, a crevasse or a firn aquifer fracturing resulting from the bending of an ice shelf as surface lakes can change the ice rheology both on grounded ice and ice shelves fill and drain32,33,74. through the release of latent heat42. Hydrofracture can occur on both floating and grounded ice. The The widespread and intense surface melt in Greenland today is process occurs when the hydrostatic pressure at the tip of a water- a template for understanding surface hydrology in Antarctica in a filled crevasse exceeds the ambient pressure sufficiently to induce warmer world. To date, the first mode(direct surface melt) is wide- stresses at the tip of the crevasse that overcome the fracture tough- spread in Greenland and on some Antarctic ice shelves14. The sec- ness. If water fills the fracture as it grows vertically, it may fracture ond mode, injection of surface water to the bed, is also widespread the full ice thickness35,36,47,68,87,88. Water can be supplied from a lake, in Greenland39,72,73 but has not yet been observed in Antarctica. The stream or firn aquifer. Whether hydrofracture triggers ice-shelf col- third mode, meltwater-induced ice-shelf collapse, has been impli- lapse will depend on the fracture spacing. Closely spaced through- cated in the widespread collapses of northeast Antarctic Peninsula ice fractures are more likely to lead to an unstable ice shelf. When ice shelves, including Larsen A and Prince Gustav in 1995, Larsen B the fractured ice-shelf fragments have aspect ratios of horizontal in 2002 and Wilkins in 200811,35,36,74. length to ice thickness that are less than a critical value (~0.6)89, iceberg capsizing can drive ice-shelf disintegration86,90. In contrast, Surface melt leading to direct surface runoff and thinning. widely spaced fractures will not lead to iceberg capsize and instead In Antarctica, the first mode (Mode 1 in Fig. 3a,b), is primarily may provide conduits to remove the surface meltwater buffering the impacting ice shelves, whereas in Greenland, surface melt plays an ice shelf from collapse1,91. important role in mass balance of the entire ice sheet. Before 2006, Ponding of surface meltwater can also trigger ice-shelf col- mass loss in Greenland was equally partitioned between losses from lapse through ice-shelf flexing, weakening and fracturing, as lakes surface melt and runoff and loss due to ice dynamics75. Beginning fill and drain74,88. An ice shelf deflects downwards when a surface in 2006, the surface melt mass loss increased, exceeding the mass lake fills, and hydrostatically rebounds upwards when a lake rap- loss attributed to ice dynamics75,76. In a recent study, up to 84% of idly drains. This loading and unloading of surface lakes can pro- the annual mass loss from the Greenland Ice Sheet was attributed to duce flexurally induced ring and radial fractures around the lake74,92, surface melt and runoff76. Surface melting and runoff have contrib- as observed around drained lakes on the Shackleton Ice Shelf uted to the lowering of the ice sheet margin77 at rates of >1 m yr−1. (Fig. 1d) and the Langhovde Ice Shelf, East Antarctica12. A chain Close to the ice-sheet margin, surface meltwater is exported directly reaction of lake drainage events could occur if these loading-induced off the ice in supraglacial streams. Inland, the surface water can fractures intersect adjacent lakes. The adjacent lakes will drain into refreeze, be stored near the surface44,78 or be transported to the base and deepen the new fracture. This chain reaction process may have of the ice sheet38,39,79. As Antarctic melt rates increase in the future, triggered the drainage of over 2,000 meltwater lakes19 in the weeks mass loss due to surface runoff will also increase. before the collapse of the Larsen B Ice Shelf74. Meltwater-induced flexure and fracture may also have contributed to the 2008 break- Injection of surface meltwater into the subglacial environment. up events of the Wilkins Ice Shelf11. Chain-reaction lake drainages The second mode of impact (Mode 2 in Fig. 3c,d) has not been will only occur if lakes are close enough that fractures formed by documented in Antarctica yet, but is widespread in Greenland. In one lake drainage event intersect an adjacent lake74. Stresses from Greenland, the surface and basal hydrological systems are linked by further afield, including back-stress from land-fast sea ice87 and drainage of surface lakes into fractures79, and drainage of surface larger-scale ice flow, can mute the impact of loading and unloading rivers into moulins80. Meltwater stored in the englacial hydrological by preventing fracture initiation. Some surface lakes have persisted system as subsurface lakes41,43 and firn aquifers44 may also move sur- on ice shelves, such as the George VI Ice Shelf, for decades without face water to the ice-sheet base47. For example, transient storage of triggering collapse14. Although George VI Ice Shelf (Fig. 1a) is cov- surface meltwater in a firn aquifer upslope of Helheim Glacier, east ered with widespread, closely spaced lakes every year, its compres- Greenland, flows downslope until it disappears at an extensional sive flow regime93 limits the formation of fractures — even with the crevasse. Modelling suggests that this water reaches the ice-sheet persistent loading from abundant surface meltwater. bed via hydrofracture47. Injection of surface water to the subglacial Most of our understanding of ice-shelf collapse comes from hydrological system may increase ice mass loss through enhanced the Antarctic Peninsula. It is likely that more Antarctic ice shelves basal sliding2 and enhanced ocean melting at calving fronts. Sudden will also be impacted by hydrofracture, as warming produces more lake drainage events can produce both localized vertical and hori- melting in tandem with sustained wind-enhanced melting, result- zontal ice displacements38,39,81,82. Together, the seasonal evolution of ing in the reduced permeability of ice shelf firn and allowing the surface meltwater, its transfer to the subglacial environment and the formation of melt ponds in vulnerable areas. efficiency of subglacial hydrological systems modulate the response of ice dynamics to meltwater input83,84. In Greenland, research has Role of meltwater in future mass balance focused on both the short-term (hours to weeks)38,39 and the sea- In the future, surface melting will play an increasingly important sonal response of the ice sheet to meltwater injections85,86 as an role in Antarctic Ice Sheet mass balance as the climate warms in analogue for understanding how the ice sheet will respond dynami- response to GHG emissions94,95. The degree of influence will cally to increased surface melt. There is currently no evidence for depend critically on melt rates, which increase nonlinearly with coupling between Antarctic surface and basal hydrological systems. atmospheric temperatures — mainly as a result of the melt–albedo

Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange 1047 Perspective NAture CliMAte ChAnGe a Incoming shortwave b

Outgoing shortwave from snow

Mode 1 Lakes Surface lowering, Reduced outgoing thinning shortwave from lakes Runoff

Runoff

Bedrock Bedrock c d Ice velocity Ice velocity

Firm Lake Lake aquifier Calving front

Mode 2 Calving front

Crevasses Moulin

Subglacial drainage Melting Ocean mixing Increased Increased Changed melting ocean mixing Bedrock Bedrock basal conditions e f

Calving front Lake Calving front Increased ice velocity

Mode 3 Fractures Ice velocity Icebergs Ice shelf Ice shelf

Grounding line Grounding line

Bedrock Bedrock

Fig. 3 | Schematic illustration of three primary modes of surface melt impact on ice-sheet mass balance. a,b, Mode 1. Direct surface ablation is enhanced over lake bottoms owing to the albedo feedback30, resulting in incoming shortwave radiation reflecting less (small yellow arrow) from lakes than adjacent snow or bare ice surfaces (larger arrow). c,d, Mode 2. Connectivity between the ice surface hydrology and ice-sheet base impacts ice dynamics by modifying basal thermal and hydrologic conditions. Connections may occur through surface lakes draining into fractures, via rivers draining into moulins and via firn aquifers draining into fractures. e,f, Mode 3. Meltwater-induced ice-shelf collapse due to presence of surface lakes. Surface lakes propagate pre-existing fractures downwards by hydrofracture (light blue lake and fracture)68,88 and load (or unload) the ice shelf, creating new fractures (dark blue lake and fractures) that drain adjacent lakes74. When an ice shelf collapses, mass loss will increase as the reduced buttressing force will trigger the outlet glaciers to accelerate.

positive feedback94. This positive feedback heightens the sensitiv- Evidence for melt–temperature nonlinearity and its impacts is pro- ity of warmer regions to future temperature increases, while also vided by an ice core on the northeastern Antarctic Peninsula, docu- enabling melt to shift from a relatively insignificant process to a menting rapid melt intensification since the mid-twentieth century potentially dominant driver of ice-shelf change over this century. coincident with numerous ice shelf collapses55.

1048 Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange NAture CliMAte ChAnGe Perspective a b

0 250 500 km 0 500 1,000 1,500 2,000 km

Surface melt (mm w.e. yr–1)

0 250 500 750 1,000 1,250

Fig. 4 | Surface meltwater production in Greenland today and Antarctica at the end of the century. a, Mean annual surface melting in Greenland as simulated over 2000–2009 in MARv3.5.2 forced by ERA-Interim100. b, Projected annual surface melting over 2091–2100 in Antarctica under RCP8.5 using an ensemble of CMIP5-based models94. Note that the colour scale is different to that in Fig. 1. Contours indicate surface elevation at intervals of 500 m.

Simulations of future Antarctic surface melting vary widely, Figure 4 compares melt rates projected for the end of the century with the dominant source of uncertainty in projections result- in Antarctica under RCP8.5 to present-day melt rates in Greenland. ing from the uncertainty in the future evolution of GHG emis- This provides a framework for understanding the future impact sions (that is, scenario uncertainty). Additional uncertainty of melt in Antarctica. The region that will experience the great- emerges from the biases inherent to various climate models, as est increase in surface melt will be the Antarctic Peninsula. Melt well as the configuration of modelling experiments and uncer- rates as high as in Greenland’s lower ablation zone, where surface tainty in the parameterization of meltwater transport, storage meltwater is connected to the bed, are projected for this region by and influence on ice-shelf fracture. Owing to the nonlinear sen- 2100. Melt intensity is strongly dependent on elevation and latitude. sitivity of melt to temperature change, even small biases in the If not already at risk of collapse due intensified surface melting94, simulation of present-day climate can translate to large biases in Antarctic Peninsula ice shelves will probably deplete their firn air the simulation of future meltwater production. Illustrating this content under a high-emissions scenario by the end of the century6. case, models that do not reproduce melt conditions today project Lack of pore space within the firn layer of Antarctic Peninsula ice 200–500% more melt by 210095 than a subset of climate models shelves will heighten their sensitivity to further melt increases by that are able to reproduce present-day melt rates94. Nevertheless, promoting meltwater pooling or runoff as opposed to percolation even under more conservative projections94, a near doubling of and refreezing45. A simplistic interpretation of this comparison sug- the Antarctic-wide volume of melt is simulated by 2050, irrespec- gests that bare ice zones, melt lakes and moulins will replace percola- tive of the emissions scenario selected. Beyond mid-century, there tion zones that proliferate across much of floating and grounded ice is a close coupling between CO2 emissions and Antarctic melt. of the Antarctic Peninsula today. This could trigger several meltwa- Under a high-emissions scenario (Representative Concentration ter impacts that are active in Greenland but are currently negligible Pathway (RCP) 8.5), melt on nearly all Antarctic Peninsula ice in Antarctica, including meltwater runoff and the injection of melt- shelves — and to a lesser degree on ice shelves further south in water to the bed. Given historical melt-rate and temperature-based West Antarctica — approaches or surpasses levels associated with thresholds for ice-shelf viability94, the Larsen C Ice Shelf and others recent Antarctic Peninsula ice-shelf collapses94. Other projections on the Antarctic Peninsula can be expected to collapse under this with more intense surface melt95 suggest that by 2100 surface melt emissions scenario this century35,53. With high melt intensification will trigger rapid and widespread Antarctic ice-sheet mass losses projected and increased snowfall already observed96, firn aquifers through a progression of instability mechanisms including surface and subsurface lakes may develop along the Antarctic Peninsula. melt-induced ice-shelf hydrofracture, marine ice-cliff instability The impact of surface hydrology on ice-sheet mass balance in and marine ice-sheet instability89. Here we will focus on the more other parts of Antarctica will grow as the extent and intensity of sur- conservative of these two model-based studies, albeit under a face melt increases. The ponding of meltwater on ice shelves could high-emissions pathway. contribute to their collapse. Whether water is exported by ice-shelf

Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange 1049 Perspective NAture CliMAte ChAnGe rivers will depend on surface slope, surface conditions and the ice- The identification of new constraints on ice structure, the evolution shelf stress state. If predictions of increased melting are accurate, and drivers of melt through time and the vulnerability of ice shelves by 2100 ice shelves in the Antarctic Peninsula will probably have to hydrofracture should include ice cores and geophysical map- collapsed and all remaining ice shelves including the large Ross, ping. Sustained and robust observations of Antarctic surface melt Filchner-Ronne and Amery ice sheets will undergo firn densifica- and hydrological processes are needed, particularly to constrain tion due to the increased surface melt. Atmospherically driven sur- their varied drivers and impacts on ice properties and stability, to face lowering due to firn compaction will be occurring in tandem then develop and refine parameterizations of these processes in with the ocean-driven basal thinning of ice shelves that is already continental-scale ice-sheet models. These are critical knowledge acting on much of peripheral Antarctica97,98. Meltwater may collect gaps that limit our understanding of future Antarctic mass change. at the grounding lines of the large ice shelves similar to the ponding Addressing these uncertainties will require a sustained, coordi- and refreezing that is occurring at the grounding line of the Larsen nated, international and interdisciplinary effort. C Ice Shelf today. The elevated surface melt on the Abbott, Getz and The impact of increased surface melting on the mass balance Shackleton ice shelves will lead to the collapse of these ice shelves of the Antarctic Ice Sheet will depend on the fate of the meltwater unless active surface drainage can mitigate the effect of surface as melt on vulnerable buttressing ice shelves increases and that on loading by exporting water to the ocean. the grounded ice begins to resemble the melt storage, transport and On the grounded portions of East and West Antarctica, surface export active today in Greenland. Whether future surface melt and lowering due to runoff and connectivity to the bed (modes 1 and 2, hydrology resembles that experienced by early Antarctic explorers, Fig. 3) could become significant by 2100 in certain regions. Regions or that found in Greenland today, is tied in large part to the future where 2100 melt rates similar to those observed in Greenland today emissions of GHGs. In the near future, surface melt processes will develop on grounded Antarctic ice include the Pine Island catch- have the greatest impact on global sea level through susceptible ice ment and portions of Wilkes Land, East Antarctica. We expect that shelves buttressing large catchments. When and where each mode areas of englacial water storage — including firn aquifers and bur- of meltwater impact — direct thinning, injection of meltwater to ied lakes — will expand as accumulation and precipitation increase the bed and hydrofracture — are activated in a wetter, warmer simultaneously this century99. Antarctica will to some extent control to how much Antarctica con- Increased snow accumulation, a result of a warming atmosphere, tributes to global sea-level rise. is likely to moderate the impact of melt. Recent coupled climate modelling indicates that owing to the enhanced moisture-holding Received: 30 March 2018; Accepted: 4 October 2018; capacity of the atmosphere and increased open-ocean evapora- Published online: 19 November 2018 tion, Antarctic surface mass balance may increase by 70 Gt yr−1 99 per degree of warming even as surface melt and runoff increase . References Evidence for ongoing warming-enhanced snowfall is preserved in 1. Bell, R. E. et al. Antarctic ice shelf potentially stabilized by export of ice cores. Increased snowfall could also inhibit the melt–albedo meltwater in surface river. Nature 544, 344–348 (2017). feedback, which is important for melt initiation and seasonal melt Presents evidence of persistent removal of surface meltwater from an ice evolution on East Antarctic ice shelves7. Enhanced snowfall may shelf by a surface river and the potential to develop surface drainage on Antarctica’s large ice shelves. also support the growth of the ice shelf/sheet firn layer and thus 2. Zwally, H. J. & Fiegles, S. Extent and duration of Antarctic surface melting. enable enhanced meltwater infiltration and refreeze as opposed to J. Glaciol. 40, 463–475 (1994). ponding45 or promote future growth in meltwater storage in aqui- 3. Liu, H., Wang, L. & Jezek, K. C. Spatiotemporal variations of snowmelt in fers18,45. If the firn layer thickens, more meltwater will infiltrate and Antarctica derived from satellite scanning multichannel microwave refreeze or be stored in firn aquifers18 rather than ponding on the ice radiometer and Special Sensor Microwave Imager data (1978–2004). 45 J. Geophys. Res. Earth Surf. 111, F01003 (2006). surface . While increased accumulation may buffer the impact of 4. Tedesco, M., Abdalati, W. & Zwally, H. J. Persistent surface snowmelt increased surface melt on runoff and ice-shelf collapse, if increased over Antarctica (1987–2006) from 19.35 GHz brightness temperatures. accumulation leads to the formation of extensive firn aquifers in Geophys. Res. Lett. 34, L18504 (2007). crevassed regions, connectivity between the surface and basal 5. Trusel, L. D., Frey, K. E. & Das, S. B. Antarctic surface melting dynamics: hydrologic systems may develop. Similarly, an increase in ice flux enhanced perspectives from radar scatterometer data. J. Geophys. Res. Earth Surf. 117, F02023 (2012). could result from meltwater injected into ice shear margins or into 6. Munneke, P. K., Ligtenberg, S. R. M., Van den Broeke, M. R. & Vaughan, D. G. regions of Antarctica with cold frozen beds. Firn air depletion as a precursor of Antarctic ice-shelf collapse. J. Glaciol. To move beyond simple projections of modern Greenland 60, 205–214 (2014). hydrology to a warmer Antarctica requires an improved under- 7. Trusel, L. D., Frey, K. E., Das, S. B., Munneke, P. K. & Broeke, M. R. Satellite‐based estimates of Antarctic surface meltwater fuxes. Geophys. Res. standing of surface hydrology on ice shelves and ice sheets. There Lett. 40, 6148–6153 (2013). are profound knowledge gaps in our understanding of the role 8. Lenaerts, J. T. M. et al. Meltwater produced by wind–albedo interaction of firn densification, the roles of hydrofracture and meltwater- stored in an East Antarctic ice shelf. Nat. Clim. Change 7, 58–62 (2017). loading-induced-flexure on ice-shelf fracture and calving, and how 9. Nicolas, J. P. et al. January 2016 extensive summer melt in West Antarctica effective surface rivers are in buffering ice shelves from collapse — favoured by strong El Niño. Nat. Commun. 8, 15799 (2017). 10. Nghiem, S. et al. Te extreme melt across the Greenland ice sheet in 2012. these must be addressed to inform our grasp of surface hydrology. Geophys. Res. Lett. 39, L20502 (2012). Similarly, for grounded ice, we do not have a clear understanding 11. Glasser, N. F. et al. Surface structure and stability of the Larsen C ice shelf, of what happens when surface melt first reaches the base of an ice Antarctic Peninsula. J. Glaciol. 55, 400–410 (2009). sheet. Because of melt–temperature nonlinearity and the varied 12. Langley, E. S., Leeson, A. A., Stokes, C. R. & Jamieson, S. S. R. Seasonal local- and global-scale processes impacting melt, it is essential for evolution of supraglacial lakes on an East Antarctic outlet glacier. Geophys. Res. Lett. 43, 8563–8571 (2016). climate and ice-sheet models to realistically simulate present-day One of the earliest detailed multiyear analyses of surpraglacial Antarctic climate. pond formation and drainage in East Antarctica, suggesting three Accurate estimates of surface meltwater production today are ways in which lakes evolve: refreeze, drainage englacially or hampered by lack of continuity in satellite datasets, and the sparse drain supraglacially. spatial and temporal in situ observations necessary to constrain the 13. Phillips, H. A. Surface meltstreams on the Amery Ice Shelf, East Antarctica. Ann. Glaciol. 27, 177–181 (1998). surface energy balance. New satellite campaigns (such as Landsat 14. Kingslake, J., Ely, J. C., Das, I. & Bell, R. E. Widespread movement 8 and the Sentinel constellation) and dedicated field campaigns in of meltwater onto and across Antarctic ice shelves. Nature 544, melt-prone areas are beginning to address this observation void. 349–352 (2017).

1050 Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange NAture CliMAte ChAnGe Perspective

Presents satellite evidence for widespread meltwater ponds, streams and 41. Miles, K. E., Willis, I. C., Benedek, C. L., Williamson, A. G. & Tedesco, M. rivers on the fanks of Antarctic Ice Sheets and surrounding ice shelves, Toward monitoring surface and subsurface lakes on the Greenland ice suggesting that active surface hydrology is more common today than sheet using Sentinel-1 SAR and Landsat-8 OLI imagery. Front. Earth Sci. 5, previously thought. 58 (2017). 15. Bindschadler, R., Scambos, T. A., Rott, H., Skvarca, P. & Vornberger, P. Ice 42. Hubbard, B. et al. Massive subsurface ice formed by refreezing of ice-shelf dolines on Larsen Ice Shelf, Antarctica. Ann. Glaciol. 34, 283–290 (2002). melt ponds. Nat. Commun. 7, 11897 (2016). 16. Humphrey, N. F., Harper, J. T. & Pfefer, W. T. Termal tracking of 43. Koenig, L. S. et al. Wintertime storage of water in buried supraglacial lakes meltwater retention in Greenland’s accumulation area. J. Geophys. Res. across the Greenland Ice Sheet. Cryosphere 9, 1333–1342 (2015). Earth Surf. 117, F01010 (2012). 44. Forster, R. R. et al. Extensive liquid meltwater storage in frn within the 17. Polashenski, C. et al. Observations of pronounced Greenland ice sheet frn Greenland ice sheet. Nat. Geosci. 7, 95–98 (2014). warming and implications for runof production. Geophys. Res. Lett. 41, 45. Munneke, P. K., M Ligtenberg, S. R., Broeke, M. R., Angelen, J. H. & 4238–4246 (2014). Forster, R. R. Explaining the presence of perennial liquid water 18. Munneke, P. K. et al. Intense winter surface melt on an Antarctic ice shelf. bodies in the frn of the Greenland Ice Sheet. Geophys. Res. Lett. 41, Geophys. Res. Lett. 45, 7615–7623 (2018). 476–483 (2014). 19. Glasser, N. F. & Scambos, T. A. A structural glaciological analysis of the 46. Miège, C. et al. Spatial extent and temporal variability of Greenland frn 2002 Larsen B ice-shelf collapse. J. Glaciol. 54, 3–16 (2008). aquifers detected by ground and airborne radars. J. Geophys. Res. Earth 20. Banwell, A. F. et al. Supraglacial lakes on the Larsen B ice shelf, Antarctica, Surf. 121, 2381–2398 (2016). and at Paakitsoq, West Greenland: a comparative study. Ann. Glaciol. 55, 47. Poinar, K. et al. Drainage of Southeast Greenland frn aquifer water through 1–8 (2014). crevasses to the bed. Front. Earth Sci. 5, 5 (2017). 21. Liston, G. E., Winther, J.-G., Bruland, O., Elvehøy, H. & Sand, K. 48. Miller, O. L. et al. Hydraulic conductivity of a frn aquifer in southeast Below-surface ice melt on the coastal Antarctic ice sheet. J. Glaciol. 45, Greenland. Front. Earth Sci. 5, 38 (2017). 273–285 (1999). 49. Noël, B. et al. Modelling the climate and surface mass balance of polar ice 22. Liston, G. E. & Winther, J.-G. Antarctic surface and subsurface snow and sheets using RACMO2–Part 1: Greenland (1958–2016). Cryosphere 12, ice melt fuxes. J. Clim. 18, 1469–1481 (2005). 811–831 (2018). 23. Winther, J.-G., Elvehøy, H., Bøggild, C. E., Sand, K. & Liston, G. Melting, 50. Kingslake, J. & Sole, A. Modelling channelized surface drainage of runof and the formation of frozen lakes in a mixed snow and blue-ice feld supraglacial lakes. J. Glaciol. 61, 185–199 (2015). in Dronning Maud Land, Antarctica. J. Glaciol. 42, 271–278 (1996). 51. Turner, J. Signifcant warming of the Antarctic winter troposphere. Science 24. Echelmeyer, K., Clarke, T. S. & Harrison, W. D. Surfcial glaciology of 311, 1914–1917 (2006). Jakobshavns Isbræ, West Greenland. Part I: surface morphology. J. Glaciol. 52. Scambos, T., Hulbe, C. & Fahnestock, M. in Antarctic Peninsula Climate 37, 368–382 (1991). Variability: A Historical and Paleoenvironmental Perspective (eds 25. Miles, B. W. J., Stokes, C. R., Vieli, A. & Cox, N. J. Rapid, climate-driven Domack, E. W. et al.) 79–92 (Antarctic Research Series 79, American changes in outlet glaciers on the Pacifc coast of East Antarctica. Nature Geophysical Union, Washington DC, 2003). 500, 563–566 (2013). 53. Vaughan, D. G. & Doake, C. S. M. Recent atmospheric warming 26. McGrath, D. et al. Basal crevasses on the Larsen C Ice Shelf, Antarctica: and retreat of ice shelves on the Antarctic Peninsula. Nature 379, implications for meltwater ponding and hydrofracture. Geophys. Res. Lett. 328–331 (1996). 39, L16504 (2012). 54. Pritchard, H. D. & Vaughan, D. G. Widespread acceleration of tidewater 27. Glasser, N. F. & Gudmundsson, G. H. Longitudinal surface structures glaciers on the Antarctic Peninsula. J. Geophys. Res. Earth Surf. 112, (fowstripes) on Antarctic glaciers. Cryosphere 6, 383–391 (2012). F03S29 (2007). 28. Dow, C. F. et al. Basal channels drive active surface hydrology and 55. Abram, N. J. et al. Acceleration of snow melt in an Antarctic Peninsula ice transverse ice shelf fracture. Sci. Adv. 4, eaao7212 (2018). core during the twentieth century. Nat. Geosci. 6, 404–411 (2013). 29. Bevan, S. L. et al. Centuries of intense surface melt on Larsen C Ice Shelf. 56. Marshall, G. J. Trends in the Southern Annular Mode from observations Cryosphere 11, 2743 (2017). and reanalyses. J. Clim. 16, 4134–4143 (2003). Using observations from fve 90-m-deep boreholes and results from a 57. Tompson, D. W. et al. Signatures of the Antarctic ozone hole in Southern fow-line model and a frn densifcation model, this study shows that Hemisphere surface climate change. Nat. Geosci. 4, 741–749 (2011). stacked layers of re-frozen pond ice correspond to two climatic warm 58. Tedesco, M. & Monaghan, A. J. An updated Antarctic melt record through periods within the last 300 years on the Antarctic Peninsula. 2009 and its linkages to high‐latitude and tropical climate variability. 30. Tedesco, M. et al. Measurement and modeling of ablation of the Geophys. Res. Lett. 36, L18502 (2009). bottom of supraglacial lakes in western Greenland. Geophys. Res. Lett. 39, 59. Deb, P. et al. Summer drivers of atmospheric variability afecting ice shelf L02502 (2012). thinning in the Amundsen Sea Embayment, West Antarctica. Geophys. Res. 31. Lüthje, M., Pedersen, L. T., Reeh, N. & Greuell, W. Modelling the evolution Lett. 45, 4124–4133 (2018). of supraglacial lakes on the West Greenland ice-sheet margin. J. Glaciol. 52, 60. Nicolas, J. P. & Bromwich, D. H. Climate of West Antarctica and infuence 608–618 (2006). of marine air intrusions. J. Clim. 24, 49–67 (2011). 32. MacAyeal, D. R., Sergienko, O. V. & Banwell, A. F. A model of viscoelastic 61. Winther, J.-G., Jespersen, M. N. & Liston, G. E. Blue-ice areas in Antarctica ice-shelf fexure. J. Glaciol. 61, 635–645 (2015). derived from NOAA AVHRR satellite data. J. Glaciol. 47, 325–334 (2001). 33. Banwell, A. F. & Macayeal, D. R. Ice-shelf fracture due to viscoelastic 62. Hui, F. et al. Mapping blue-ice areas in Antarctica using ETM+ and MODIS fexure stress induced by fll/drain cycles of supraglacial lakes. Antarct. Sci. data. Ann. Glaciol. 55, 129–137 (2014). 27, 587–597 (2015). 63. Luckman, A. et al. Surface melt and ponding on Larsen C Ice Shelf and the 34. Fricker, H. A. et al. Mapping the grounding zone of the Amery Ice Shelf, impact of foehn winds. Antarct. Sci. 26, 625–635 (2014). East Antarctica using InSAR, MODIS and ICESat. Antarct. Sci. 21, 64. Turton, J. V., Kirchgaessner, A., Ross, A. N. & King, J. C. Does high‐ 515–532 (2009). resolution modelling improve the spatial analysis of föhn fow over the 35. Scambos, T. et al. Ice shelf disintegration by plate bending and hydro- Larsen C Ice Shelf? Weather 72, 192–196 (2017). fracture: satellite observations and model results of the 2008 Wilkins ice 65. Barrett, J. E. et al. Persistent efects of a discrete warming event on a polar shelf break-ups. Earth Planet. Sci. Lett. 280, 51–60 (2009). desert ecosystem. Glob. Change Biol. 14, 2249–2261 (2008). 36. Scambos, T. A., Hulbe, C., Fahnestock, M. & Bohlander, J. Te link between 66. Speirs, J. C., Steinhof, D. F., McGowan, H. A., Bromwich, D. H. & climate warming and breakup of ice shelves in the Antarctic Peninsula. Monaghan, A. J. Foehn winds in the McMurdo Dry Valleys, Antarctica: the J. Glaciol. 46, 516–530 (2000). origin of extreme warming events. J. Clim. 23, 3577–3598 (2010). Outlines concept that crevasse propagation by meltwater is a key 67. Stern, A. A., Dinniman, M. S., Zagorodnov, V., Tyler, S. W. & Holland, D. M. mechanism that contributes to the break up of Antarctic Pennisula ice Intrusion of warm surface water beneath the McMurdo ice shelf, shelves. Antarctica. J. Geophys. Res. Oceans 118, 7036–7048 (2013). 37. Macdonald, G. J., Banwell, A. F. & MacAyeal, D. R. Seasonal evolution of 68. van der Veen, C. J. Fracture propagation as means of rapidly transferring supraglacial lakes on a foating ice tongue, Petermann Glacier, Greenland. surface meltwater to the base of glaciers. Geophys. Res. Lett. 34, Ann. Glaciol. 59, 56–65 (2018). L01501 (2007). 38. Das, S. B. et al. Fracture propagation to the base of the Greenland Ice Sheet 69. Cape, M. R. et al. Foehn winds link climate‐driven warming to ice shelf during supraglacial lake drainage. Science 320, 778–781 (2008). evolution in Antarctica. J. Geophys. Res. Atmos. 120, 11037–11057 (2015). 39. Tedesco, M. et al. Ice dynamic response to two modes of surface 70. Straneo, F. et al. Impact of ford dynamics and glacial runof on the lake drainage on the Greenland ice sheet. Environ. Res. Lett. 8, circulation near Helheim Glacier. Nat. Geosci. 4, 322–327 (2011). 034007 (2013). 71. Fried, M. J. et al. Distributed subglacial discharge drives signifcant 40. Paige, R. A. Sub-surface melt pools in the McMurdo Ice Shelf, Antarctica. submarine melt at a Greenland tidewater glacier. Geophys. Res. Lett. 42, J. Glaciol. 7, 511–516 (1968). 9328–9336 (2015).

Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange 1051 Perspective NAture CliMAte ChAnGe

72. Bartholomew, I. et al. Supraglacial forcing of subglacial drainage in 94. Trusel, L. D. et al. Divergent trajectories of Antarctic surface melt the ablation zone of the Greenland ice sheet. Geophys. Res. Lett. 38, under two twenty-frst-century climate scenarios. Nat. Geosci. 8, L08502 (2011). 927–932 (2015). 73. Hofman, M., Catania, G., Neumann, T., Andrews, L. & Rumrill, J. Links Shows that surface melting in Antarctica grows exponentially in between acceleration, melting, and supraglacial lake drainage of the western response to a warming atmosphere, and that by 2100 under a high- Greenland Ice Sheet. J. Geophys. Res. Earth Surf. 116, F04035 (2011). emissions scenario, multiple ice shelves exceed levels of melt associated 74. Banwell, A. F., MacAyeal, D. R. & Sergienko, O. V. Breakup of the Larsen B with previous ice-shelf collapses. Ice Shelf triggered by chain reaction drainage of supraglacial lakes. Geophys. 95. DeConto, R. M. & Pollard, D. Contribution of Antarctica to past and future Res. Lett. 40, 5872–5876 (2013). sea-level rise. Nature 531, 591–597 (2016). Shows that a surface lake drainage event on an ice shelf may produce Includes new parameterizations for ice-shelf hydrofracture and ice-clif fractures that intersect other lakes, initiating the self-stimulating lake failure in an ice sheet–climate model, suggesting that atmospheric drainage mechanism that the authors propose may have contributed to warming may drive Antarctic mass losses, both in previous warm the widespread break-up of the Larsen B Ice Shelf in 2002. periods and in the twenty-frst century and beyond. 75. van den Broeke, M. et al. Partitioning recent Greenland mass loss. Science 96. Tomas, E. R., Marshall, G. J. & McConnell, J. R. A doubling in snow 326, 984–986 (2009). accumulation in the western Antarctic Peninsula since 1850. Geophys. Res. 76. Enderlin, E. M. et al. An improved mass budget for the Greenland ice Lett. 35, L01706 (2008). sheet. Geophys. Res. Lett. 41, 866–872 (2014). 97. Pritchard, H. D. et al. Antarctic ice-sheet loss driven by basal melting of ice 77. Csatho, B. M. et al. Laser altimetry reveals complex pattern of Greenland shelves. Nature 484, 502–505 (2012). Ice Sheet dynamics. Proc. Natl Acad. Sci. USA 111, 18478–18483 (2014). 98. Paolo, F. S., Fricker, H. A. & Padman, L. Volume loss from Antarctic ice 78. Cooper, M. G. et al. Meltwater storage in low-density near-surface bare ice shelves is accelerating. Science 348, 327–331 (2015). in the Greenland ice sheet ablation zone. Cryosphere 12, 955–970 (2018). 99. Lenaerts, J. T. M., Vizcaino, M., Fyke, J., Van Kampenhout, L. & van den 79. Willis, M. J., Herried, B. G., Bevis, M. G. & Bell, R. E. Recharge of a Broeke, M. R. Present-day and future Antarctic ice sheet climate subglacial lake by surface meltwater in northeast Greenland. Nature 518, and surface mass balance in the Community Earth System Model. 223–227 (2015). Clim. Dynam. 47, 1367–1381 (2016). 80. Smith, L. C. et al. Efcient meltwater drainage through supraglacial streams Shows that surface melting is enhanced near the grounding zone of at and rivers on the southwest Greenland ice sheet. Proc. Natl Acad. Sci. USA least one East Antarctic ice shelf due to positive wind–albedo–melt 112, 1001–1006 (2015). feedbacks, suggesting the potential for ice-shelf hydrofracture beyond 81. Stevens, L. A. et al. Greenland supraglacial lake drainages triggered by the Antarctic Peninsula. hydrologically induced basal slip. Nature 522, 73–76 (2015). 100. Fettweis, X. et al. Reconstructions of the 1900–2015 Greenland ice sheet 82. Joughin, I. Seasonal speedup along the western fank of the Greenland Ice surface mass balance using the regional climate MAR model. Cryosphere Sheet. Science 320, 781–783 (2008). 11, 1015 (2017). 83. Sundal, A. V. et al. Melt-induced speed-up of Greenland ice sheet ofset by efcient subglacial drainage. Nature 469, 521–524 (2011). Acknowledgements 84. Tedstone, A. J. et al. Greenland ice sheet motion insensitive to exceptional A.F.B. acknowledges support from a Leverhulme/Newton Trust Early Career meltwater forcing. Proc. Natl Acad. Sci. USA 110, 19719–19724 (2013). Fellowship (grant no. ECF-2014-412). We thank X. Fettweis for producing and 85. Hofman, M. J. et al. Widespread moulin formation during supraglacial lake making Greenland MAR model output available and M. Tedesco for discussion drainages in Greenland. Geophys. Res. Lett. 45, 778–788 (2018). on their usage. The authors thank O. Sergienko for useful discussions. The authors 86. Banwell, A., Hewitt, I., Willis, I. & Arnold, N. Moulin density controls acknowledge the participants in the February 2018 NSF-funded workshop on Antarctic drainage development beneath the Greenland ice sheet. J. Geophys. Res. Surface Hydrology and Future Ice Shelf Stability (grant no. 1743326) for their lively, Earth Surf. 121, 2248–2269 (2016). thoughtful discussion. L.D.T. acknowledges support from NSF Antarctic Glaciology 87. Van der Veen, C. J. Fracture mechanics approach to penetration of surface Program award no. 1643733. crevasses on glaciers. Cold Reg. Sci. Technol. 27, 31–47 (1998). 88. MacAyeal, D. R. & Sergienko, O. V. Te fexural dynamics of melting ice shelves. Ann. Glaciol. 54, 1–10 (2013). Author contributions 89. Burton, J. C. et al. Laboratory investigations of iceberg capsize dynamics, R.E.B. conceived the idea, and all authors contributed equally to the writing. energy dissipation and tsunamigenesis. J. Geophys. Res. Earth Surf. 117, F01007 (2012). Competing interests 90. MacAyeal, D. R., Scambos, T. A., Hulbe, C. L. & Fahnestock, M. A. The authors declare no competing interests. Catastrophic ice-shelf break-up by an ice-shelf-fragment-capsize mechanism. J. Glaciol. 49, 22–36 (2003). 91. Banwell, A. Glaciology: ice-shelf stability questioned. Nature 544, Additional information 306–307 (2017). Reprints and permissions information is available at www.nature.com/reprints. 92. Beltaos, S. Collapse of foating ice covers under vertical loads: test data vs. Correspondence should be addressed to R.E.B. theory. Cold Reg. Sci. Technol. 34, 191–207 (2002). 93. Hambrey, M. J. et al. Structure and sedimentology of George VI Ice Shelf, Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in Antarctic Peninsula: implications for ice-sheet dynamics and landform published maps and institutional affiliations. development. J. Geol. Soc. 172, 599–613 (2015). © Springer Nature Limited 2018

1052 Nature Climate Change | VOL 8 | DECEMBER 2018 | 1044–1052 | www.nature.com/natureclimatechange